Top 5 vector databases and when to use them

Vector databases are the future for semantic search, similarity search, clustering, and recommendations for both text and images. Here’s how and when to use them.

Photo by Jan Antonin Kolar on Unsplash

Vector databases have been the hot new thing in the database space for a while now. Even PostgreSQL has added an extension, pgvector, with support for vector fields and cosine similarity search. Vector databases power a laundry list of use cases for your business:

• Fast, reasonably accurate search through semantic search. Use the right model, and you can even get multilingual search for cheap.
• Text content clustering.
• Image search and image similarity search.
• Text and image recommendations without the cold-start problem.

This is accomplished through the magic of embeddings, which convert a text or an image into a high-dimensional vector. In this article, we will be working with text — if you would like me to do a writeup of how image embeddings work, then leave a comment.

This gives us the opportunity to work with text and images using the vast array of numerical tools that computers excel at. Most importantly, we can use what is known as cosine…